Interfacing a 12-bit temperature sensor with ATMEL controller

[Boncuk]

Hi all,

I plan to build an air conditioning device for the entire house with each room controlling the temperature individually.

Since temperature differences of 0.5 degrees are already considered as a big difference by people I want to make control with an accuracy of 0.1 degrees. ATMEL controllers have A/D converters, but they just have a resolution of 10 bits which would result in an accuracy of 0.5 degrees.

Has anyone tried to interface a precision temperature sensor 12 bit or more to the ATMEL?

If so, I'd appreciate some hints how to manage that.

The MCU just has to manage one analog input and will return two logical outputs, one for cooling and the other one for heating (if necessary) and of course, an LC-display, one UP, one DOWN and one ENTER key.

Regards

Hans

 

[donniedj]

Since temperature differences of 0.5 degrees are already considered as a big difference by people I want to make control with an accuracy of 0.1 degrees.

That sounds a bit excessive. 72.5 to 73.0 is not a big difference. Perhaps .5 degrees Celsius which would result in 1 degree of change in Fahrenheit. That can be felt but is still not a big difference. A cooling system cannot and needs not to have .1 or .5 degrees of control for simple domestic use because a room is too large and too dynamic. A computer monitor could go to sleep, a light turn on or off, even a persons mood would change the ambient so quickly that your system will not respond in time which ends up overcompensating. The room will always have its own microclimate also. Ankle level will be cooler, neck level hotter, and west will differ from the east. This is even true for an igloo.

0.5 degrees more than suffices.

 

[blueroomelectronics]

Most household thermostats are off by +/-2C anyway. You don't need 0.1F precision for a home HVAC system.

 

[Oznog]

A thermistor isn't even this accurate, not without calibration.
DS18S20 Dallas 1-wire is a nice accurate device and you get a 1-wire network here.
I agree there's no benefit to under 1deg F of accuracy, or even 2 deg F really. In fact there's no way to have a room temp with great accuracy- the area by an outside wall is a degree or two cooler. When the heat kicks on the air will heat up faster than the thermistor. Air near the TV is a few deg warmer. Air currents near the heater vent are 5degF warmer when the heater's on. Then the heat turns off and the air is warmer than the objects IN the room. So, which temp do you want here?

Besides, the stability would be wacked. The heat turns on to the room, shuts off at 68.0F, then you shut off the heater but the blower keeps running for like a minute after the gas shuts off and it gets up to 68.1. Then just a couple of minutes later the temp's down to 59.9F and you fire up the heater again? We gonna keep pulsing that whole furnace on and off to maintain meaningless temp accuracies?

 

[Salgat]

Hysteresis exists for a reason. Constantly turning on and shutting off A/C creates unnecessary stress on the unit.

 

[simrantogether]

Well...

Hi,

Simply use DS1820 or LM34 ( for farenheit ) or LM35 ( For Degree Celcius ).

Regards,

Simran..

 

[eblc1388]

I don't understand what's the problem with 10-bit AD.

The usable range is 0~1023, which can mean 0~102.3 degree or roughly +/- 50.0 degree with resolution of 0.1 degree.

You only need to condition the analogue input signal level to fall into this range and all is well.

 

[Boncuk]

Come on friends,

that's not what I wanted to know. All of you assume normal distribution of cool or warm air.

My professional main fields are air flow control and aerodynamics. So I know how to evenly distribute air masses with exactly the same temperature even in big size rooms as production halls. Concerning air pressure my goal was always to achieve constant pressure of +/- 1Pa, which worked in almost every case.

Living in Thailand does not necessarily mean that I share the lack of accuracy with domestic people here.

Assuming the control circuit can handle an accuracy of +/-2 deg C I'd rip it off for the junk yard, since the actual inaccuracy will reach about 3 degrees, far beyond the temperature range people makes feel comfortable, hot or cold.

All I wanted to know which kind of sensor would be easy to interface with an Atmel MCU. It can't be the Dallas type since it is already inaccurate from design.

If you are satisfied with what you have I won't object. For me it's not perfect.

Regards

Hans

 

[Nigel Goodwin]

My professional main fields are air flow control and aerodynamics. So I know how to evenly distribute air masses with exactly the same temperature even in big size rooms as production halls. Concerning air pressure my goal was always to achieve constant pressure of +/- 1Pa, which worked in almost every case.

I would imagine if you were 'professional' you wouldn't make such a ludicrous claim? - 'exactly' the same temperature throughout a large hall?, I don't think so!.

 

[Diver300]

The big advantage of a good temperature resolution is that you can see small changes, even if the temperature is not that accurate.

I would suggest the LM95071 from national. Accuracy is +/- 1 deg C and the resolution is 1/32 deg C. That would let you show 0.1 deg F steps if you wanted.

**broken link removed**

The interface is 3 wires, Serial Clock, Serial Data and Not Chip Select. You don't have to worry about anything analogue.

You program the microcontroller to lower the Not Chip Select. Then the logic level on the Serial Data is the first bit of the temperature. Each clock cycle results in the next least significant temperature bit, and you can stop at any time by raising Not Chip Select.

 

[Boncuk]

I would imagine if you were 'professional' you wouldn't make such a ludicrous claim? - 'exactly' the same temperature throughout a large hall?, I don't think so!.

Hi Nigel,

as you certainly know 'ludicrous' also stands for 'laughable', 'absurd', 'exaggerated' and 'eccentric'.

There is none of those applicable from my point of view.

Of course using radiators for equal heat distribution is the worst approach to success. You are free to get prove of my 'claim' at Woods of Colchester Ltd.

High precision is possible in every field of physics. If the engineers concerned with satellites would do a sloppy job not considering the required precision the satellites would either fall back on earth (and thereby evaporate) or disappear in space instead of orbiting.

Anyway, your post didn't even come close to the thread.

Hans

 

[blueroomelectronics]

You're dealing with humans, what feels hot to one may feel cool to others. As long as your thermostat results are repeatable most people use a form of fuzzy logic to adjust it to what feels comfortable. Your HVAC doesn't like to be tweaked, generally you turn it off / on once every 10 minutes. That's why they have anticipators.
https://home.howstuffworks.com/home-thermostat2.htm

 

[Boncuk]

You're dealing with humans, what feels hot to one may feel cool to others.

That's exactly the reason why each room has its own temperature sensor and controller. Controlling either temperature or air pressure the controller must look ahead and control according to tendencies too. Turning off either heating or cooling at the desired temperature is already too late. The controller must cut off and also engergize early enough for accuracy.

Compare it with an autopilot. Climbing to a desired flight level at a climb rate of 2.500ft/min the pitch angle has to be reduced at least 1.000 ft before reaching the desired altitude. If maintained constant until reaching it the overshoot will be approx. 700 ft, far off tolerances granted by air traffic control. (tolerance allowed 200ft off assigned altitude)

Since my air conditioning device doesn't use a medium cooled below the freezing point (what a waste!) to mix with ambient air the control will be very smooth and accurate if accurate sensors are used too. Starting out with a sensor tolerance of +/- 1 degree the controller won't do the job properly, just the same as I posted in another thread. <gabage in = garbage out>.

 

[Boncuk]

Hi Diver300,

thanks for the advice. I downloaded the data sheet and the sensor date look very promising.

With an accuracy of 0.03125 deg/C it is more than what I expected.

Regards

Boncuk

 

[BeeBop]

Yes, but did you catch what Salgat was trying to tell you?

If your air conditioner compressor has been 'pumped up' it will need two to three minutes off time, to bleed out, before you start it again. Trying to start in a loaded condition often will wear out your motor in no time.

 

[Boncuk]

Yes, but did you catch what Salgat was trying to tell you?

If your air conditioner compressor has been 'pumped up' it will need two to three minutes off time, to bleed out, before you start it again. Trying to start in a loaded condition often will wear out your motor in no time.

Of course I caught. However I'm not going to waste immense energy using a compressor. One degree of cooling takes double the money of heating. I prefer applying the so called 'chill effect' using simply water. I won't also use the method applied in the USA blowing damp air into the house. Instead I will use two water circulation units, one for the primary water and another for the secondary.

The primary water circuit will be blown through the system with twin axial fans in parallel with three speeds and simultaneously activate spray nozzles working at 100bar pressure causing fog in the chamber for optimum cooling. Controlling that circuit comparing the ambient temperature and the desired room temperature already has the effect of mimimum energy to be applied. E.g. ambient temperature 28 degC, desired room temperature 24 degC there won't be any reason to run two relatively big sized fans (1m dia each) at full speed and inject water from six nozzles.

Of course, using a compressor it will cool rapidly, but on the other hand consume a lot of electric power mixing a medium cooled below freezing point with ambient air. So why should go that way and mix "supercool" air with ambient to achieve comfortable temperatures, which doesn't apply either. Most people using that kind of air conditioning run around the entire year with a cold. Starting out inaccurate accuracy will never be achieved.

Using compressors for air cooling is similar to shooting little birds with an artillery gun.

Boncuk

 

[Boncuk]

I don't understand what's the problem with 10-bit AD.

The usable range is 0~1023, which can mean 0~102.3 degree or roughly +/- 50.0 degree with resolution of 0.1 degree.

You only need to condition the analogue input signal level to fall into this range and all is well.

If there weren't a math error in your calculation it would be fine to use. The resolution is 1 degree.

 

[Pommie]

If there weren't a math error in your calculation it would be fine to use. The resolution is 1 degree.

I don't see any error. Care to elaborate?

Mike.

 

[Boncuk]

I don't see any error. Care to elaborate?

Mike.

No elaboration necessary. The error was on my side. Please accept my apologies.

Regards

Hans